sign in sign up
<<
Home , Arabian partridge, Bamboo partridge, Chinese bamboo partridge, Congo peafowl, Francolinus, Green junglefowl

Molecular Phylogenetics and Vol. 11, No. 1, February, pp. 38–54, 1999 Article ID mpev.1998.0562, available online at http://www.idealibrary.com on

A Molecular Phylogeny of the and Suggests That These Lineages Are Not Monophyletic R. T. Kimball,* E. L. Braun,*,† P. W. Zwartjes,* T. M. Crowe,‡,§ and J. D. Ligon* *Department of Biology, University of New Mexico, Albuquerque, New Mexico 87131; †National Center for Genome Resources, 1800 Old Pecos Trail, Santa Fe, New Mexico 87505; ‡Percy FitzPatrick Institute, University of Capetown, Rondebosch, 7700, South Africa; and §Department of , American Museum of Natural History, Central Park West at 79th Street, New York, New York 10024-5192

Received October 8, 1997; revised June 2, 1998

World partridges are smaller and widely distributed in Cytochrome b and D-loop nucleotide sequences were Asia, Africa, and Europe. Most are used to study patterns of molecular evolution and monochromatic and primarily dull colored. None exhib- phylogenetic relationships between the pheasants and its the extreme or highly specialized ornamentation the partridges, which are thought to form two closely characteristic of the pheasants. related monophyletic galliform lineages. Our analyses Although the is well defined, taxo- used 34 complete cytochrome b and 22 partial D-loop nomic relationships are less clear within the group sequences from the hypervariable domain I of the (Verheyen, 1956), due to the low variability in anatomi- D-loop, representing 20 species (15 genera) and 12 partridge species (5 genera). We performed cal and osteological traits (Blanchard, 1857, cited in parsimony, maximum likelihood, and distance analy- Verheyen, 1956; Lowe, 1938; Delacour, 1977). In addi- ses to resolve these phylogenetic relationships. In this tion to the study of anatomical traits (e.g., Verheyen, data set, transversion analyses gave results similar to 1956), other traits such as tail molt patterns (Beebe, those of global analyses. All of our molecular phyloge- 1914) or combinations of morphological and behavioral netic analyses indicated that the pheasants and par- traits (e.g., Delacour, 1977) also have been employed in tridges arose through a rapid radiation, making it attempts to ascertain relationships within the order. difficult to establish higher level relationships. How- Johnsgard (1986, 1988) and Sibley and Ahlquist (1990) ever, we were able to establish six major lineages provide detailed reviews of galliform systematics and containing pheasant and partridge taxa, including one the relationships among the pheasants and partridges. lineage containing both pheasants and partridges (Gal- Johnsgard (1986, 1988) concludes that the pheasants lus, Bambusicola and ). This result, sup- and partridges probably form two monophyletic lin- ported by maximum likelihood tests, indicated that the eages in the subfamily (Fig. 1A). Using pheasants and partridges do not form independent DNA hybridization, Sibley and Ahlquist (1990) also monophyletic lineages. ௠ 1999 Academic Press indicate that both the pheasants and the partridges are monophyletic. Johnsgard (1986) suggests that the pheasants evolved from a generalized partridge-like INTRODUCTION ancestor and that the early radiation of the partridge and pheasant lineages probably occurred in southeast The pheasants and Old World partridges are thought Asia. Four major pheasant lineages are recognized by to represent two closely related taxa within the order Johnsgard (1986): (1) the gallopheasants and their Galliformes (tribes Phasianini and Perdicini, respec- allies; (2) the and their allies; (3) the tively; Johnsgard, 1986, 1988). The pheasants are and their allies; and (4) the (Fig. 1B). relatively large with most species exhibiting Johnsgard (1988) also constructed a dendrogram of the extreme sexual dichromatism. Typically, male pheas- partridge genera, but considered it highly speculative. ants are brightly colored and have well developed Akishinonomiya et al. (1995) sequenced the hypervari- ornamental traits such as elongated tails, crests, and able domain I of the D-loop (mitochondrial control specialized fleshy structures. Even monochromatic spe- region) to examine relationships both among pheasant cies of pheasants exhibit some degree of ornamenta- taxa and between pheasants and partridges. Although tion. Pheasants are confined to Asia, except for the Akishinonomiya et al. (1995) examined species from (Afropavo congensis), which has a re- only three of Johnsgard’s (1986) four proposed pheas- stricted distribution in Africa. In contrast, the Old ant lineages, his results provide some support for these

38 1055-7903/99 $30.00 Copyright ௠ 1999 by Academic Press All rights of reproduction in any form reserved. MOLECULAR PHYLOGENY OF THE PHEASANTS AND PARTRIDGES 39

FIG. 1. Johnsgard’s (1986) hypothesized relationships among (A) Galliformes and (B) the pheasants. major lineages. Unfortunately, possibly due to limited MATERIALS AND METHODS taxon sampling, the data presented by Akishinonomiya et al. (1995) does not resolve the relationships within Molecular Biology Techniques the partridges or the relationship between the pheas- We extracted DNA from blood or tissue (breast muscle) ants and the partridges. Akishinonomiya et al. (1995) and amplified the cytochrome b gene by PCR using stan- did note a high degree of uncorrected sequence identity dard protocols described elsewhere (Kimball et al., 1997). between the partridge (Bambusicola) and mem- Sequencing reactions were performed as described previ- bers of the junglefowl (Gallus) and peafowl () ously by Kimball et al. (1997) or using the Thermo-Sequen- genera, leading those authors to suggest tentatively ase dye terminator kit (Amersham) according to the manu- that the ancestor of Bambusicola may also have been facturer’s recommendations. The primers used for both the ancestor of a lineage that evolved into the Gallus PCR amplification and sequencing are listed in Table 1. and Pavo clades. However, the reliability of this result was not examined, and no data were provided to indicate whether similar results are found when more TABLE 1 sophisticated methods of phylogenetic analysis are Amplification and Sequencing Primers employed. Moreover, neither the basal members of the for Cytochrome b Pavo clade (Argusianus and Polyplectron) nor any other pheasant examined show a high degree of similar- Namea Sequence (5Ј = 3Ј) Source ity to Bambusicola. L14731 ATCGCCTCCCACCT(AG)AT(CG)GA This study In this paper, we present phylogenetic analyses L14851 TACCTGGGTTCCTTCGCCCT Kornegay et al., 1993 based upon complete DNA sequences of the mitochon- L14990 ATCCAACATCTCAGCATGATGAAA Modified, Kornegay drial cytochrome b gene from all but one monospecific et al., 1993 pheasant genus, including representatives of each pro- L15164 GCAAACGGCGCCTCATTCTT This study H15298 CCTCAGAATGATATTTGTCCTCA Modified, Kornegay posed major lineage, as well as several partridge gen- et al., 1993 era. We used the molecular data to examine hypotheses L15311 CTCCCATGAGGCCAAATATC Modified, Kornegay of the evolution of the pheasants and partridges, focus- et al., 1993 ing on evolutionary relationships: (1) among the pheas- H15400 AGGGTTGGGTTGTCGACTGA This study ants; (2) between the pheasants and the partridges; L15662 CTAGGCGACCCAGAAAACTT This study H15670 GGGTTACTAGTGGGTTTGC This study and (3) with other galliforms. We also reexamined L15737 CCTATTTGCTTACGCCATCCT This study hypervariable domain I D-loop sequences from galli- H15826 CGGAAGGTTATGGTTCGTTGTTT This study forms (Akishinonomiya et al., 1995; Kimball et al., H16065 TTCAGTTTTTGGTTTACAAGAC Modified, Kornegay 1997; Lopez et al., unpublished GenBank submissions) et al., 1993 to assess the congruence of estimates of the phylogeny a Names indicate light (L) or heavy (H) strand and the position of obtained using this region of the mitochondrial genome the 3Ј end of the oligonucleotide numbered according to the with those obtained using cytochrome b. mitochondrion (Desjardins and Morais, 1990). 40 KIMBALL ET AL.

Primers designed for this study were based upon galliform bp), so alignment was straightforward. D-loop domain I sequence data. sequences were aligned using the default parameters Southern hybridization was conducted using stan- in ClustalW (Thompson et al., 1994), followed with dard methods (Ausubel et al., 1994). Briefly, selected optimization by eye. Regions with many gaps were DNA samples (see Table 2) were digested using EcoRI, removed from analyses (see Table 3). The D-loop se- separated by agarose gel electrophoresis, transferred to quence of Francolinus had many unresolved bases Hybond Nϩ (Amersham) under alkaline conditions, (Kimball et al., 1997), and removing these sites left and hybridized in 50% formamide buffer to a segment fewer sites for analysis. Therefore, most D-loop analy- of cytochrome b corresponding to the region amplified ses excluded Francolinus. We deleted all unresolved from Gallus gallus using primers L15662 and H16065 sites for analyses that included Francolinus. 32 and labeled with P. Phylogenetic Analyses Sequence Alignment and Taxon Selection Maximum parsimony analyses (unweighted parsi- The species we examined are listed in Table 2. Avian mony and transversion parsimony) were performed cytochrome b sequences are uniform in length (1143 using PAUP 3.1.1 (Swofford, 1993). Constraint trees

TABLE 2

Species Examined and Source of Sequence Data

Group Species Common name Cyt. ba D-loopb

Cracids Ortalis vetula Plain Chachalaca L08384 — Crax pauxi Helmeted Currassow AF068190 — Turkeys Meleagris gallopavo L08381 — Tympanchus phasianellus Sharp-tailed Grouse AF068191 — Numida meleagris L08383 AF013765 gambelii Gambel’s Quail L08382 — montezumae Montezuma Quail AF068192 — Partridges barbara Barbary Partridge Z48771 Y08556 Alectoris chukar Chukar L08378 D66890 Alectoris graeca Z48772 Z80942 Alectoris magna Przevalski’s Partridge Z48776 — Alectoris melanocephala Arabian Partridge Z48773 — Alectoris philbyi Philby’s Partridge Z48774 — Alectoris rufa Red-legged Partridge Z48775 Y08555 Bambusicola thoracica Chinese AF028790 D66889 coturnix Japanese Quail L08377 D82924 Coturnix sinensis Blue-breasted Quail — D66888 Francolinus francolinusc Black AF013762 AF013766 perdix Grey Partridge AF028791 D66891 Pheasants Afropavo congensis Congo Peafowl AF013760 AF013764 Argusianus Great Argus Pheasant AF013761 D66898 Catreus wallichic AF028792 — pictusc AF028793 D66895 Crossoptilon crossoptilonc White- AF028794 — Gallus gallus /Chicken AF028795 X52392 Gallus lafayettei Junglefowl — D66893 Gallus sonnerati — D66892 Gallus varius — D64163 Ithaginis cruentus Blood Pheasant AF068193 — Lophophorus impejanus Himalayan AF028796 — Lophura nycthemera L08380 D66897 Pavo cristatus L08379 D66900 Pavo muticus AF013763 D64164 colchicusc Ring-neck Pheasant AF028798 D66894d Polyplectron bicalcaratumc Gray Peacock–Pheasant AF028799 D66899 Pucrasia macrolopha AF028800 — humiae Mrs. Hume’s Pheasant — D66896 Syrmaticus reevesic Reeve’s Pheasant AF028801 — temminckiic Temminck’s Tragopan AF028802 —

a Cytochrome b sequences from this study and Kornegay et al., 1993; Randi, 1996; and Kimball et al., 1997. b D-loop sequences from Desjardins and Morais, 1990; Akishinonomiya et al., 1995; Kimball et al., 1997; and Lopez et al., unpublished GenBank submission. c Examined using Southern blot analysis. d Akishinonomiya et al., 1995 lists D66894 as Phasianus colchicus, while the database entry lists D66894 as Phasianus versicolor. MOLECULAR PHYLOGENY OF THE PHEASANTS AND PARTRIDGES 41 were constructed using MacClade 3.05 (Maddison and Justification of these parameters is presented under Maddison, 1992) and then the most parsimonious trees Results. Comparison of alternative phylogenetic trees given the constraints were identified using PAUP 3.1.1 was performed using the test proposed by Kishino and (Swofford, 1993). Parsimony analyses used at least 100 Hasegawa (1989) as implemented in DNAml. random addition sequence replicates and the following Distance analyses were performed using PHYLIP settings: TBR branch swapping, collapse yes, mulpars (Felsenstein, 1993) and MOLPHY (Adachi and Hase- yes, steepest descent no. We used MacClade 3.05 gawa, 1996). We used the K2P (Kimura 2-parameter) ϩ (Maddison and Maddison, 1992) to reconstruct charac- ␥ (Jin and Nei, 1990) and F84 (Kishino and Hasegawa, ter evolution using parsimony. 1989) models of DNA sequence evolution, since these The reliability of specific groupings in parsimony models accommodate site-to-site rate heterogeneity trees was assessed using the bootstrap (Felsenstein, and unequal nucleotide frequencies, respectively. Dis- 1985). We estimated the bootstrap proportion in parsi- tance estimates were calculated using transition– mony analyses using 1000 replicates, with 10 random transversion ratios of 4 and 10 for cytochrome b, and addition sequence replicates for each bootstrap repli- K2P ϩ␥distance estimates were computed using a cate. A number of studies have suggested that the coefficient of variation of 2.24 (which is equivalent to bootstrap proportion is a conservative estimator of the ␣ϭ0.2). Protein distances were calculated from trans- probability that a clade is correct, as long as the method lated cytochrome b sequences using both ProtML (Ada- used to estimate phylogenetic relationships is consis- chi and Hasegawa, 1996) with options -D (distance tent (Hillis and Bull, 1993; Rodrigo et al., 1994). In fact, matrix) and -mf (mtREV24 with empirical amino acid several studies have suggested that for maximum frequencies, as described by Adachi and Hasegawa, parsimony bootstrap values Ն70%, the probability of a 1995) and ProtDist (Felsenstein, 1993) with the PAM clade being correct is at least 95% (Hillis and Bull, model of evolution (Dayhoff et al., 1978). Trees were 1993), although some authors have questioned whether inferred from distance matrices using neighbor joining accepting monophyly of a group whose bootstrap propor- (Saitou and Nei, 1987). tion is relatively low (around 70–80%) in the absence of prior expectation might inflate type I error (Rodrigo et al., 1994). We feel that clades showing less than 50% RESULTS bootstrap in all analyses are unreliable and we have Molecular Evolution of Cytochrome b collapsed these in nucleotide analyses. We consider and D-loop Sequences clades to be well supported when the bootstrap propor- tion is Ն70% and the clade is present in multiple data All cytochrome b sequences contained an open read- sets. ing frame that encoded a protein with significant Maximum likelihood estimation was performed us- identity to other cytochrome b proteins. The heme- ing DNAml (Felsenstein, 1993) or PUZZLE (Strimmer ligating histidines and other conserved residues (How- and von Haeseler, 1997) using either the F84 (described ell, 1989) could be identified, suggesting that our by Kishino and Hasegawa, 1989) or the HKY85 (Ha- sequences were functional cytochrome b genes, rather segawa et al., 1985) models of DNA sequence evolution. than nuclear pseudogenes (e.g., Kornegay et al., 1993; We accommodated site-to-site rate heterogeneity using Arctander, 1995). An analysis of nuclear pseudogenes a four-category discrete approximation of a ␥ distribu- and their functional counterparts by Sorenson and tion (Yang, 1994) with ␣ϭ0.2 estimated by maximum Quinn (1998) indicated that nuclear pseudogenes often likelihood using PUZZLE (␣ϭ0.2 corresponds to four accumulate mutations that would result in amino acid equiprobable categories with relative rates of 0.0002, changes in highly conserved regions even in the ab- 0.0382, 0.4882, and 3.4733). To determine whether the sence of indels (e.g., Arctander, 1995), making an incorporation of site-to-site rate heterogeneity resulted examination of such regions a suitable method to detect in a significant improvement in the model, we used the nuclear pseudogenes. likelihood ratio test and compared the test statistic We examined several other lines of evidence to 2 (␦ϭ2[ln L1 Ϫ ln L2]) to the ␹ distribution with one determine whether the sequences we analyzed repre- degree of freedom (corresponding to the addition of the sented nuclear pseudogenes (see Sorenson and Quinn, shape parameter of the ␥ distribution). This test ap- 1998). Sequences for two species that we analyzed pears to be robust as long as the number of parameters (Phasianus colchicus and Polyplectron bicalcaratum) represented by the different models is clear, as it is in were confirmed using mitochondrially enriched tissues. this case (Huelsenbeck et al., 1996). For maximum Southern blot analysis of eight species (see Table 2) likelihood analyses of cytochrome b sequences, we demonstrated that only one restriction fragment - conducted six random addition sequence replicates in ized with a cytochrome b probe. Branch lengths of DNAml using a transition–transversion ratio of 10, nuclear pseudogenes tend to be shorter than their rate categories corresponding to a discrete approxima- functional counterparts (Sorenson and Fleischer, 1996; tion to a ␥ distribution with ␣ϭ0.2, and the global Sorenson and Quinn, 1998). An examination of cyto- rearrangements option (75,802 trees were examined). chrome b phylograms (e.g., Fig. 2) suggests that se- 42 KIMBALL ET AL.

FIG. 2. Most likely tree identified using cytochrome b nucleotide data. In likelihood ϭϪ11698.3. quences obtained from blood samples were not associ- codon positions. The 1143 bp cytochrome b alignment ated with short branch lengths. contained 527 variable sites, of which 422 were informa- As previously observed (e.g., Kornegay et al., 1993), tive (parsimony) sites. Most of the variable sites were in the base composition of the cytochrome b sequences the third position of codons, with 359 variable and 321 was highly biased, with the strongest bias in third- informative third-codon positions. In the translated MOLECULAR PHYLOGENY OF THE PHEASANTS AND PARTRIDGES 43 data set, 113 amino acids were variable, and 62 amino be superior for the estimation of deeper branches. acid positions were informative. However, trees estimated using either method are We identified 16 equally parsimonious trees by un- largely congruent (Fig. 3), suggesting that the differ- weighted parsimony analysis of the cytochrome b nucle- ences between the transition and the transversion data otide alignment (tree length ϭ 2427; CI excluding partitions are fairly modest. Parsimony analyses were uninformative sites ϭ 0.307). Previous analyses of mi- largely congruent with distance analyses (unpublished tochondrial genes, including cytochrome b, have indi- data); thus we did not present the results of distance cated that analysis of transversions may improve pholo- analyses. genetic reconstruction, particularly at deeper branches Previous analyses have suggested that mitochon- (e.g., Mindell and Thacker, 1996). Therefore, we per- drial protein coding sequences exhibit substantial site- formed transversion parsimony and identified a total of to-site rate heterogeneity (Kumar, 1996). Our previous 66 equally parsimonious trees (tree length ϭ 675; CI analysis of cytochrome b sequences from a more limited excluding uninformative sites ϭ 0.381). These analyses set of galliforms indicates that these sequences do indicate that there is less homoplasy in the transver- exhibit substantial site-to-site rate heterogeneity (Kim- sion data, suggesting that transversion analyses may ball et al., 1997). We found that incorporating site-to-

FIG. 3. Bootstrap consensus tree of cytochrome b nucleotide sequences. Numbers are percentage bootstrap support for unweighted parsimony (above branch) and transversion parsimony (below). No data are given if bootstrap values are Ͻ50%. 44 KIMBALL ET AL. site rate heterogeneity resulted in a significant improve- 306 bases of which 96 were variable and 60 were ment in the estimates of likelihood (ln L ϭϪ11698.25 informative. [␣ϭ0.2]; ln L ϭϪ13625.14 [no rate heterogeneity]; Estimates of phylogeny obtained using domain I of ␦ϭ1926.88, significant at P Ͻ 0.001). Based upon these the D-loop (Fig. 5) are largely congruent with those results, we used a discrete approximation of a ␥ distri- obtained using cytochrome b (compare Figs. 3 and 5), bution with ␣ϭ0.2 for estimation of phylogenies by although there are differences in the taxon composition maximum likelihood. of these two data sets. The congruence between these Previous analyses have also indicated that mitochon- data sets may seem surprising since the hypervariable drial sequences show an extremely high transition– domain I of the D-loop is generally thought to be transversion ratio (Wakeley, 1996). However, estima- inadequate for the analysis of mid- to deep-level tion of the transition–transversion ratio is not branches in avian phylogeny due to problems with completely straightforward, and it has been reported saturation. Furthermore, the fact that substantially that maximum likelihood methods underestimate the fewer sites were available for phylogenetic analyses of transition–transversion ratio of mitochondrial se- the hypervariable domain I of the D-loop (350 bp) than quences (Purvis and Bromham, 1997). For this reason, there were for phylogenetic analyses of the complete the maximum likelihood estimate of the transition– cytochrome b gene (1143 bp) suggests that the D-loop transversion ratio for the galliform cytochrome b data analyses would present additional problems. However, analyzed in this study, which corresponds to 3.8, prob- the similarity between the estimates of phylogeny ably represents a minimum value for the actual transi- obtained using domain I of the D-loop and cytochrome b tion–transversion ratio. In fact, pairwise estimates of suggests that the D-loop alignment contained substan- the transition–transversion ratio, calculated using a tial phylogenetic information. K2P ϩ␥correction for multiple substitutions (Jin and These results are consistent with recent simulation Nei, 1990), range from 1.2 to 18.4, with comparisons studies that suggest that rapidly evolving sequences between more closely related taxa consistently corre- may actually have extremely desirable properties for sponding to the higher estimates of the transition– phylogenetic reconstruction, even for divergent taxa transversion ratio. For this reason, we have used a (Hillis, 1998; Yang, 1998). It is possible to find empirical transition–transversion ratio of 10, as have several support for these simulations, such as the study of previous phylogenetic analyses of avian cytochrome b Lewis et al. (1997), which showed that accurate phylog- enies of liverworts that diverged over 400 million years sequences (e.g., Nunn and Cracraft, 1996; Nunn et al., ago could be inferred using only third-codon positions 1996; Kimball et al., 1997). from rbcL sequences, despite the high degree of diver- Phylogenetic analyses of inferred cytochrome b amino gence at these sites. Unweighted parsimony analysis of acid sequences were largely congruent with estimates the galliform taxa for which both cytochrome b and of phylogeny based upon cytochrome b nucleotide se- D-loop sequences were available (see Table 2) indicate quences (Fig. 4). However, as previous studies of galli- that there are similar levels of homoplasy in the form cytochrome b sequences have indicated (Kornegay alignments of cytochrome b and domain I of the D-loop. et al., 1993; Randi, 1996; Kimball et al., 1997), the For these taxa, we identified a single most parsimoni- bootstrap support for most specific groupings was ex- ous tree using the cytochrome b alignment (tree tremely low. Previous studies have shown that the use length ϭ 1169; CI excluding uninformative sites ϭ of amino acid sequences rather than nucleotide se- 0.440) and two most parsimonious trees using the quences for phylogenetic analyses of mitochondrial domain I D-loop alignment (tree length ϭ 268; CI ex- protein coding genes may discard more information cluding uninformative sites ϭ 0.480). Based upon these than noise (Milinkovitch et al., 1996), which may reflect results, we feel that analyses of D-loop sequences show the functional constraints upon the proteins encoded by potential for resolution of avian phylogenies at multiple the mitochondrial genome (see Naylor et al., 1995). levels. We conducted phylogenetic analyses of the hypervari- able domain I of the D-loop using galliform sequence data available from the National Center for Biotechnol- Relationships within the Pheasants ogy Information (Table 2). After alignment of these The four major lineages of pheasants (Fig. 1B) pro- sequences and elimination of regions where homology posed by Johnsgard (1986) are largely supported in at of nucleotides was ambiguous (see Table 3), we were least some of our analyses. However, within the lin- left with a 350-bp alignment of D-loop sequences that eages, we inferred different branching orders from exhibited a somewhat biased nucleotide composition those proposed by Johnsgard (1986). In the junglefowl (see Kimball et al., 1997). This alignment contained 115 lineage, our analyses suggested the inclusion of addi- variable sites and 79 informative sites. Inclusion of tional genera not previously suggested to be pheasants. Francolinus and deletion of sites that are ambiguous in Monophyly of the gallopheasant lineage is well sup- the Francolinus sequence resulted in an alignment of ported by cytochrome b bootstrap analyses (Fig. 3), and MOLECULAR PHYLOGENY OF THE PHEASANTS AND PARTRIDGES 45

FIG. 4. Analysis of cytochrome b amino acid sequences. Numbers are percentage bootstrap support for the mtREV24 model (above branch) and the PAM model (below) of evolution. No data are given if bootstrap values are Ͻ50%. is also present in the most likely tree (Fig. 2). Analysis tron are problematic. Some analyses of cytochrome b of inferred cytochrome b amino acid sequences strongly placed Polyplectron in the peafowl clade, but could not supported a clade containing most members of this resolve the position of Argusianus (e.g., Fig. 2 and group, but support for inclusion of the Lophura species distance analyses). Previous analyses of the D-loop examined is weak (Fig. 4). While a previous analysis of provided some support for the inclusion of Argusianus D-loop nucleotide sequences supported Lophura as a and Polyplectron within the peafowl clade (Akishi- member of the gallopheasant clade (Akishinonomiya et nonomiya et al., 1995; Kimball et al., 1997). Analysis of al., 1995), our current reanalysis did not support the cytochrome b protein sequences placed both taxa within inclusion of Lophura within the gallopheasant clade the peafowl clade, though bootstrap support was weak (Fig. 5). (Fig. 4). Our results supported the peafowl clade proposed by The tragopan clade proposed by Johnsgard (1986) is Johnsgard (1986), although the results are not com- the least well supported by our data. Analysis of pletely straightforward. All of our analyses strongly cytochrome b protein sequences weakly supports the supported an Afropavo–Pavo clade (also see Kimball et inclusion of Tragopan, Pucrasia, Lophophorus, and al., 1997), but the positions of Argusianus and Polyplec- Ithaginis in a clade (Fig. 4), though branching order 46 KIMBALL ET AL.

TABLE 3

Alignment of Sequences of the Hypervariable Domain I of the D-loop (Mitochondrial Control Region) from Members of the Galliformes

differs from that proposed by Johnsgard (1986; see Fig. low support for the existence of this clade suggests that 1B). Maximum likelihood analysis suggests that Trago- if these genera actually do form a clade, their diver- pan and Pucrasia form a clade, to the exclusion of gence took place relatively early in the evolution of the Lophophorus and Ithaginis (Fig. 2). Nucleotide analy- pheasants. ses cannot resolve the phylogenetic position of any of The biggest difference between our results and the the members of this hypothesized clade (Fig. 3). The lineages proposed by Johnsgard (1986) are in the MOLECULAR PHYLOGENY OF THE PHEASANTS AND PARTRIDGES 47

TABLE 3—Continued

junglefowl clade. Johnsgard (1986) suggested that the ence of a Gallus–Bambusicola clade (Fig. 5), whether or junglefowl clade contains only members of the genus not Francolinus was included in the alignment. How- Gallus (Fig. 1B). However, cytochrome b nucleotide ever, the position of Francolinus could not be resolved analyses supported a relationship between Gallus and using the reduced D-loop nucleotide data. Removal of partridges of the genera Bambusicola and Francolinus Francolinus from the cytochrome b data set resulted in (Fig. 3). Maximum likelihood and protein sequence high support for a Gallus–Bambusicola clade, indicat- analyses also suggested the presence of this clade (Figs. ing that our results are not dependent upon inclusion of 2 and 4). D-loop analyses strongly supported the pres- Francolinus (unpublished observation). 48 KIMBALL ET AL.

TABLE 3—Continued

Relationships between the Pheasants (Fig. 5). We did support monophyly of the genus Cotur- and the Partridges nix, unlike Akishinonomiya et al. (1995). Increased We analyzed relatively few partridge genera. Like taxon sampling or the analysis of the pheasant and Randi (1996), we supported monophyly of the Alectoris partridge taxa together may explain the differences partridges (Figs. 2–5) and the presence of a Coturnix– between our results and those of Akishinonomiya et al. Alectoris clade (Figs. 2–4). However, the relationship (1995). None of the analyses could resolve the position between Coturnix and Alectoris could not be resolved in of Perdix, which appears to be distantly related to all analyses of the hypervariable domain I of the D-loop other taxa sampled (Figs. 2–5). MOLECULAR PHYLOGENY OF THE PHEASANTS AND PARTRIDGES 49

TABLE 3—Continued

The available data cannot resolve the branching nus, Alectoris and Coturnix, and a final lineage contain- order of major pheasant and partridge lineages, or ing Perdix. determine whether many or all of the typical pheasant Pheasant Monophyly Can Be Excluded Based upon lineages evolved from a partridge-like ancestor as proposed by Johnsgard (1986). Instead, our data sug- Cytochrome b Sequences gest that the pheasants and partridges we sampled There are several alternative explanations for the form at least six lineages: peafowl, gallopheasants, surprising relationship between Gallus and two par- tragopans, junglefowl with Bambusicola and Francoli- tridge genera, Francolinus and Bambusicola. First, 50 KIMBALL ET AL.

TABLE 3—Continued

Note. The sequences correspond to positions 3 to 487 of the chicken mitochondrial genome (Desjardins and Morais, 1990). Dots indicate identity to the Gallus gallus sequence, dashes indicate gaps, and gray areas indicate regions deleted from phylogenetic analysis. monophyly of the pheasants and partridges may exist, ant and partridge lineages was constrained. Trees but the resolution of our data is insufficient to discrimi- identified by maximum parsimony were not signifi- nate between monophyly and polyphyly. Alternatively, cantly less likely than the most likely tree (unpublished the lineages are monophyletic, but placement of either observations). However, the most parsimonious trees Gallus or both Bambusicola and Francolinus is incor- compatible with monophyly of either the pheasants rect. To test these hypotheses, we found the most alone or the pheasants and partridges were signifi- parsimonious trees in which composition of the pheas- cantly less likely than the most likely tree (Table 4). MOLECULAR PHYLOGENY OF THE PHEASANTS AND PARTRIDGES 51

Including Gallus within the partridges or placing both TABLE 4 Bambusicola and Francolinus within the pheasants also produced trees that were significantly less likely Analysis of Cytochrome b Phylogenies in Which the Monophyly of Certain Taxa Was Constrained, Com- than the most likely tree (Table 3). Our rejection of pared with the Most Likely Tree (Fig. 2) these alternative hypotheses suggests that the lineages are clearly not monophyletic. Constrained lineages ␦ ln likelihood (SD)

Relationships with Other Galliforms Most likely tree 0.0a Ϫ Johnsgard (1986) and others place the turkeys and Pheasants and partridges 108.5 (27.8)* Pheasants only Ϫ100.3 (25.9)* grouse into two separate lineages, allied with the Gallus in Partridges Ϫ69.5 (22.2)* pheasants and partridges. Sibley and Ahlquist (1990) Bambusicola and Francolinus in Pheasants Ϫ56.3 (19.0)* suggest that turkeys are closely related to grouse, and include the turkey–grouse lineage in the pheasant and Note. Except as noted, both pheasant and partridge monophyly partridge family. Support for a turkey–grouse lineage is constrained to follow Johnsgard (1986). a ln likelihood ϭϪ11698.3. found in the most likely tree (Fig. 2) and is weakly * Significantly different from most likely tree. supported in bootstrap analyses of nucleotide data (Fig. 3). However, the clade was not present in analysis of protein sequence data. While our data cannot resolve differs from morphological (e.g., Verheyen, 1956; Johns- whether or not turkeys and grouse form a clade, these gard, 1986) and allozyme (Randi et al., 1991) analyses species are not placed outside the pheasant–partridge which place the New World quail within the pheasant– clade, suggesting that turkey and grouse evolved dur- partridge radiation. ing the radiation of the pheasants and partridges and not prior to them as has been previously suggested (e.g., Johnsgard, 1986). DISCUSSION Our analyses of cytochrome b sequences suggest that Our results showed that two genera of partridges are guineafowl and New World quail diverged prior to the present in a clade with Gallus, indicating that the radiation of the pheasants and partridges (Figs. 2–4). pheasant and partridge lineages proposed by Johns- This conclusion is congruent with some molecular gard (1986) cannot be monophyletic. Using DNA hybrid- analyses of galliform evolution (Sibley and Ahlquist, ization, Sibley and Ahlquist (1990) reported monophyly 1990; Kornegay et al., 1993; Kimball et al., 1997), but of pheasants and partridges. Although they sampled fewer taxa, their study did include Gallus and two species of Francolinus. However, they did not examine F. francolinus and the genus Francolinus is probably not monophyletic (e.g., Crowe and Crowe, 1985; Bloomer and Crowe, 1998; Laskowski and Fitch, 1989). This surprising result cannot be due to contamina- tion. The cytochrome b data for Gallus, Bambusicola, and Francolinus were collected in our lab, while the D-loop samples for Gallus were from the chicken mito- chondrion (Desjardins and Morais, 1990), and Bambu- sicola was from a study by Akishinonomiya et al. (1995). Therefore, data from different labs, sequencing different regions of mitochondrial DNA from different individuals, led to the same conclusion. In addition, other analyses have suggested a relation- ship between these taxa, though these studies did not assess the reliability of this clade. Analysis of partial ovomucoid sequences resulted in a clade containing Gallus, Bambusicola, and two Francolinus species (F. francolinus and F. pondicerianus; Laskowski and Fitch, 1989), though other Francolinus species in that study were not placed within the Gallus clade. Furthermore, a phenetic analysis of morphological data from 22 FIG. 5. Bootstrap consensus tree of D-loop nucleotide sequences. species of Francolinus and a number of other partridge Numbers are percentage bootstrap support for unweighted parsi- genera revealed that the nearest neighbor of F. lathami mony (above branch) and transversion parsimony (below). No data are given if bootstrap values are Ͻ50%. The branch labeled a was not and F. sephaena is Bambusicola, rather than the other supported by transversion parsimony. Instead, transversion parsi- species of Francolinus examined (Crowe and Crowe, mony supported a Phasianus–Syrmaticus clade at 51%. 1985). 52 KIMBALL ET AL.

The inclusion of Bambusicola and Francolinus with improve evolutionary reconstruction (Hendy and Penny, Gallus, as well as the unresolved relationship between 1989; Hillis, 1996). It is likely that such measures will the pheasants and partridges, suggests that the terms at least resolve the position of some taxa that are pheasant and partridge are not phylogenetically useful. weakly or inconsistently placed in a clade, such as Delacour (1977: 25) had noted this as well, stating ‘‘In a Polyplectron and Argusianus. strictly scientific sense, the term ‘pheasant’ applies to a Some traits may not reflect phylogenetic history group of birds which do not differ from others by when taxa undergo a rapid radiation. Several mecha- very well-defined or important characteristics . . . many nisms have been proposed to explain why this phenom- birds of these groups [partridges and quail] differ from enon may occur. Introgression of genes may occur prior one another just as much as they do from some of the to the evolution of effective isolating mechanisms, or so-called ‘pheasants’, among which, in turn, fairly polymorphisms present in the last common ancestor of distantly related genera have usually been placed.’’ It these lineages may sort randomly into multiple lin- appears that the terms pheasant and partridge should eages yielding a pattern inconsistent with the phyloge- only be used to include suites of related behavioral and netic history (reviewed in Maddison, 1996). In addition, morphological characteristics, rather than implying traits may ‘‘flicker’’ on and off during a radiation event, anything about the evolutionary history of galliform since genetic information may not degrade irretriev- birds. ably for up to 6 million years (Marshall et al., 1994). For example, genes affecting sexual selection or male orna- Our results suggest that traits generally associated mentation may have turned on and off multiple times with pheasants, such as a high degree of dichromatism in the early evolution of the pheasant and partridge and exclusive female parental care, evolved multiple lineages and may therefore not accurately reflect phylo- times within the galliforms. Members of Francolinus genetic history. We speculate that these effects may and Bambusicola are generally monochromatic, with have made it difficult to establish relationships among no highly dimorphic or ornamented species. This con- the galliforms (e.g., Delacour, 1977). trasts with the four species in the genus Gallus, all of Our results indicate that the evolution of the galli- which exhibit a high degree of ornamentation in males. forms is complex and suggest that it may be difficult to Behaviorally, Francolinus and Bambusicola are also understand the evolution of the interesting morphologi- like typical partridges, primarily exhibiting monogamy, cal, behavioral, and ecological characteristics of this while in Gallus, males often are polygynous and gener- group. However, better resolution of the deeper ally do not participate in parental care. branches, by use of additional taxa and sequence data, Support for the lability of traits typically associated may allow the reconstruction of at least some of the with pheasants can be found in the gallopheasant clade evolutionary pathways. Use of labile traits to assist in as well. Crossoptilon and Catreus are both monochro- classification of these taxa appears to have led to matic and monogamous, and they exhibit biparental misleading results, such as placing Gallus in a clade care. Their derived position in the clade suggests that with the other ‘‘pheasant-like’’ taxa, and the phyloge- they evolved from a highly dichromatic, ‘‘pheasant’’-like netic utility of such traits should be questioned. ancestor and subsequently lost dichromatism. Interest- ingly, the loss of dichromatism has differed in each lineage. In Crossoptilon, the sexes are alike, both ACKNOWLEDGMENTS exhibiting ornamentation such as ear tufts and elabo- We thank S. Lattimer and L. Krassnitzer for generously allowing rated tails; Catreus exhibits in us to collect blood samples from their birds, and E. Kempf for which the sexes are dull in coloration, but males have providing us with samples of Bambusicola. We thank J. Groth elongated tails. These different patterns suggest that (American Museum of Natural History) and M. P. Skupski (National Crossoptilon and Catreus independently evolved mono- Center for Genome Resources) for expert technical assistance. We also thank the Albuquerque High Performance Computing Center chromatism. and the National Center for Genome Resources for access to their The difficulty of resolving the branching order among computer facilities. DNA sequencing was performed in the NSF- the major galliform lineages suggests that these birds supported (RIMI Program Grant HRD-9550649) Molecular Biology underwent a relatively rapid radiation (also see Facility at the University of New Mexico. Two anonymous reviewers Kornegay et al., 1993; Kimball et al., 1997). Consistent provided helpful comments on an earlier version of the manuscript. with rapid speciation is the low bootstrap support (Figs. 3 and 5) and short branch lengths separating the REFERENCES major lineages (Figs. 2 and 4). Resolving clades and branching orders during a radiation such as this may Adachi, J., and Hasegawa, M. (1996). MOLPHY: Programs for be difficult, as there is insufficient time for genetic or Molecular Phylogenetics, ver. 2.3. Institute of Statistical Mathemat- morphological changes to accumulate before additional ics, Tokyo. Akishinonomiya, F., Miyake, T., Takada, M., Ohno, S., and Kondo, N. branching occurs. However, the addition of more se- (1995). The genetic link between the quence data for all taxa and the addition of data from (Bambusicola thoracica) and the chicken and of the more species thought to belong to these lineages may genus Gallus. Proc. Natl. Acad. Sci. USA 92: 11053–11056. MOLECULAR PHYLOGENY OF THE PHEASANTS AND PARTRIDGES 53

Arctander, P. (1995). Comparison of a mitochondrial gene and a Kumar, S. (1996). Patterns of nucleotide substitution in mitochon- corresponding nuclear pseudogene. Proc. R. Soc. London B 262: drial protein coding genes of vertebrates. Genetics 143: 537–548. 13–19. Laskowski, M. J., and Fitch, W. M. (1989). Evolution of avian Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seidman, ovomucoids and of birds. In ‘‘The Hierarchy of Life: Molecules and J. G., Smith, J. A., and Struhl, K. (1994). ‘‘Current Protocols in Morphology in Phylogenetic Analysis’’ (B. Fernholm, K. Bremer, Molecular Biology,’’ Wiley, New York. and H. Jornvall, Eds.), pp. 371–387. Elsevier, Amsterdam. Beebe, C. W. (1914). Preliminary pheasant studies. Zoologica 1: Lewis, L. A., Mishler, B. D., and Vilgalys, R. (1997). Phylogenetic 261–285. relationships of the liverworts (Hepaticae), a basal embryophyte Bloomer, P., and Crowe, T. M. (1998). Francolin phylogenetics: lineage, inferred from nucleotide sequence data of the chloroplast Molecular, morphobehavioral, and combined evidence. Mol. Phylo- gene rbcL. Mol. Phylogenet. Evol. 7: 377–393. genet. Evol. 9: 236–254. Lowe, P. R. (1938). Some preliminary notes on the and Crowe, T. M., and Crowe, A. A. (1985). The genus Francolinus as a systematic position of Afropavo congensis Chapin. In ‘‘Proceedings model for avian evolution and biogeography in Africa: 1. Relation- of the 9th International Ornithological Congress’’ (J. Delacour, ships among species. In ‘‘Proceedings of the International Ed.), pp. 219–230. Rouen. Symposium on African Vertebrates: Systematics, Phylogeny and Maddison, W. P. (1996). Molecular approaches and the growth of Evolutionary Ecology’’ (K.-L. Schuchmann, Ed.), Zoologisches phylogenetic biology. In ‘‘Molecular Zoology: Advances, Strategies, Forschungsinstitut und Museum Alexander Koenig, Bonn. and Protocols’’ (J. D. Ferraris and S. R. Palumbi, Eds.), pp. 47–63. Dayhoff, M. O., Schwartz, R. M., and Orcutt, B. C. (1978). A model of Wiley–Liss, New York. evolutionary change in proteins. In ‘‘Atlas of Protein Sequence and Maddison, W. P., and Maddison, D. R. (1992). MacClade, ver. 3.05. Structure’’ (M. O. Dayhoff, Ed.), pp. 345–352. National Biomedical Sinauer Associates, Inc., Sunderland, MA. Research Foundation, Silver Springs, MD. Marshall, C. R., Raff, E. C., and Raff, R. A. (1994). Dollo’s law and the Delacour, J. (1977). ‘‘The Pheasants of the World,’’ Saiga Pub. Co., death and resurrection of genes. Proc. Natl. Acad. Sci. USA 91: Ltd, Surrey, England. 12283–12287. Desjardins, P., and Morais, R. (1990). Sequence and gene organiza- Milinkovitch, M. C., LeDuc, R. G., Adachi, J., Farnir, F., Georges, M., tion of the chicken mitochondrial genome: A novel gene order in and Hasegawa, M. (1996). Effects of character weighting and higher vertebrates. J. Mol. Biol. 212: 599–634. species sampling on phylogeny reconstruction: A case study based Felsenstein, J. (1985). Confidence limits on phylogenies: An approach on DNA sequence data in Cetaceans. Genetics 144: 1817–1833. using the bootstrap. Evolution 39: 783–791. Mindell, D. P., and Thacker, C. E. (1996). Rates of molecular Felsenstein, J. (1993). PHYLIP (Phylogeny Inference Package), ver. evolution: Phylogenetic issues and applications. Annu. Rev. Ecol. 3.5c. Department of Genetics, University of Washington, Seattle, Syst. 27: 279–303. WA. Naylor, G. J. P., Collins, T. M., and Brown, W. M. (1995). Hydrophobic- Hasegawa, M., Kishino, H., and Yano, T. (1985). Dating of the ity and phylogeny. Nature 373: 565–566. human–ape splitting by a molecular clock of mitochondrial DNA. J. Mol. Evol. 22: 160–174. Nunn, G. B., and Cracraft, J. (1996). Phylogenetic relationships Hendy, M. D., and Penny, D. (1989). A framework for the quantitative among the major lineages of the birds-of-paradise (Paradisaeidae) study of evolutionary trees. Syst. Zool. 38: 297–309. using mitochondrial DNA gene sequences. Mol. Phylogenet. Evol. 5: 445–459. Hillis, D. M. (1996). Inferring complex phylogenies. Nature 383: 130–131. Nunn, G. B., Cooper, J., Jouventin, P., Robertson, C. J. R., and Robertson, G. G. (1996). Evolutionary relationships among extant Hillis, D. M. (1998). Taxonomic sampling, phylogenetic accuracy, and albatrosses (Procellariiformes: Diomedeidae) established from com- investigator bias. Syst. Biol. 47: 3–8. plete cytochrome b gene sequences. Auk 113: 784–801. Hillis, D. M., and Bull, J. J. (1993). An empirical test of bootstrapping Purvis, A., and Bromham, L. (1997). Estimating the transition/ as a method for assessing confidence in phylogenetic analysis. Syst. transversion ratio from independent pairwise comparisons with an Biol. 42: 182–192. assumed phylogeny. J. Mol. Evol. 44: 112–119. Howell, N. (1989). Evolutionary conservation of protein regions in the protonmotive cytochrome b and their possible roles in redox Randi, E. (1996). A mitochondrial cytochrome b phylogeny of the catalysis. J. Mol. Evol. 29: 157–169. Alectoris partridges. Mol. Phylogenet. Evol. 6: 214–227. Huelsenbeck, J. P., Hillis, D. M., and Nielsen, R. (1996). A likelihood- Randi, E., Fusco, G., Lorenzini, R., and Crowe, T. M. (1991). Phyloge- ratio test of monophyly. Syst. Biol. 45: 546–558. netic relationships and rates of allozyme evolution within the . Biochem. Syst. Ecol. 19: 213–221. Jin, L., and Nei, M. (1990). Limitations of the evolutionary parsimony method of phylogenetic analysis. Mol. Biol. Evol. 7: 82–102. Rodrigo, A. G., Bergquist, P. R., and Bergquist, P. L. (1994). Inad- equate support for an evolutionary link between the metazoa and Johnsgard, P. A. (1986). ‘‘The Pheasants of the World,’’ Oxford Univ. fungi. Syst. Biol. 43: 578–584. Press, Oxford. Saitou, N., and Nei, M. (1987). The neighbor-joining method: A new Johnsgard, P. A. (1988). ‘‘The Quails, Partridges, and of method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4: the World,’’ Oxford Univ. Press, Oxford. 406–425. Kimball, R. T., Braun, E. L., and Ligon, J. D. (1997). Resolution of the phylogenetic position of the Congo peafowl, Afropavo congensis: A Sibley, C. G., and Ahlquist, J. E. (1990). ‘‘Phylogeny and Classification biogeographic and evolutionary enigma. Proc. R. Soc. London B of Birds: A Study in Molecular Evolution,’’ Yale Univ. Press, New 264: 1517–1523. Haven, CT. Kishino, H., and Hasegawa, M. (1989). Evaluation of the maximum Sorenson, M. D., and Fleischer, R. C. (1996). Multiple independent likelihood estimate of the evolutionary tree topologies from DNA transpositions of mitochondrial DNA control region sequences to sequence data, and the branching order in Hominoidea. J. Mol. the nucleus. Proc. Natl. Acad. Sci. USA 93: 15239–15243. Evol. 29: 170–179. Sorenson, M. D., and Quinn, T. W. (1998). Numts: A challenge for Kornegay, J. R., Kocher, T. D., Williams, L. A., and Wilson, A. C. avian systematics and population biology. Auk 115: 214–221. (1993). Pathways of lysozyme evolution inferred from the se- Strimmer, K., and von Haeseler, A. (1997). PUZZLE, ver. 3.1. quences of cytochrome b in birds. J. Mol. Evol. 37: 367–379. Maximum Likelihood Analysis for Nucleotide and Amino Acid 54 KIMBALL ET AL.

Alignments. Zoologisches Institut, Universitat Munchen, Munchen, Verheyen, R. (1956). Contribution de l’anatomie et a la systematique Germany. des Galliformes. Bull. Inst. R. Sci. Nat. Belgique 32: 1–24. Swofford, D. L. (1993). PAUP: Phylogenetic Analysis Using Parsi- Wakeley, J. (1996). The excess of transitions among nucleotide mony. Formerly distributed by Illinois Natural History Survey, substitutions: New methods of estimating transition bias under- Champaign, IL. score its significance. Trends Ecol. Evol. 11: 158–163. Thompson, J. D., Higgins, D. G., and Gibson, T. J. (1994). CLUSTAL Yang, Z. (1994). Maximum likelihood phylogenetic estimation from W: Improving the sensitivity of progressive multiple sequence DNA sequences with variable rates over sites: Approximate meth- alignment through sequence weighting, position-specific gap ods. J. Mol. Evol. 39: 306–314. penalties and weight matrix choice. Nucleic Acids Res. 22: 4673– Yang, Z. (1998). On the best evolutionary rate for phylogenetic 4680. analysis. Syst. Biol. 47: 125–133.